The Within-Subject Application of Diffusion Tensor MRI and CLARITY Reveals Brain Structural Changes in Nrxn2 Deletion Mice Eleftheria Pervolaraki1†, Adam L

Pervolaraki et al. Molecular Autism (2019) 10:8 https://doi.org/10.1186/s13229-019-0261-9

RESEARCH Open Access The within-subject application of diffusion tensor MRI and CLARITY reveals brain structural changes in Nrxn2 deletion mice Eleftheria Pervolaraki1†, Adam L. Tyson2,3,4†, Francesca Pibiri5, Steven L. Poulter5, Amy C. Reichelt6, R. John Rodgers7, Steven J. Clapcote1, Colin Lever5, Laura C. Andreae2,3† and James Dachtler1,5*†

Abstract Background: Of the many genetic mutations known to increase the risk of autism spectrum disorder, a large proportion cluster upon synaptic proteins. One such family of presynaptic proteins are the neurexins (NRXN), and recent genetic and mouse evidence has suggested a causative role for NRXN2 in generating altered social behaviours. Autism has been conceptualised as a disorder of atypical connectivity, yet how single-gene mutations affect such connectivity remains under-explored. To attempt to address this, we have developed a quantitative analysis of microstructure and structural connectivity leveraging diffusion tensor MRI (DTI) with high-resolution 3D imaging in optically cleared (CLARITY) brain tissue in the same mouse, applied here to the Nrxn2α knockout (KO) model. Methods: Fixed brains of Nrxn2α KO mice underwent DTI using 9.4 T MRI, and diffusion properties of socially relevant brain regions were quantified. The same tissue was then subjected to CLARITY to immunolabel axons and cell bodies, which were also quantified. Results: DTI revealed increases in fractional anisotropy in the amygdala (including the basolateral nuclei), the anterior cingulate cortex, the orbitofrontal cortex and the hippocampus. Axial diffusivity of the anterior cingulate cortex and orbitofrontal cortex was significantly increased in Nrxn2α KO mice, as were tracts between the amygdala and the orbitofrontal cortex. Using CLARITY, we find significantly altered axonal orientation in the amygdala, orbitofrontal cortex and the anterior cingulate cortex, which was unrelated to cell density. Conclusions: Our findings demonstrate that deleting a single neurexin gene (Nrxn2α) induces atypical structural connectivity within socially relevant brain regions. More generally, our combined within-subject DTI and CLARITY approach presents a new, more sensitive method of revealing hitherto undetectable differences in the autistic brain. Keywords: MRI, CLARITY, Social, Autism, Axons, Diffusion, Structure, Imaging

* Correspondence: [email protected] †Eleftheria Pervolaraki, Adam L. Tyson, Laura C. Andreae and James Dachtler contributed equally to this work. 1School of Biomedical Sciences, University of Leeds, Leeds LS2 9JT, UK 5Department of Psychology, Durham University, South Road, Durham DH1 3LE, UK Full list of author information is available at the end of the article

© The Author(s). 2019 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Pervolaraki et al. Molecular Autism (2019) 10:8 Page 2 of 13

Background Currently, it is unknown whether deletion of Nrxn2α Autism is a common neurodevelopmental disorder, which changes the brain’s microstructure and connectivity. One is highly heritable [1]. While heritability is high, it is also previous study found coarse alterations to cell layer thick- clear that autism is highly polygenic. Around ~ 400–1000 ness within the hippocampus of Nrxn2α homozygous genes are involved in autism susceptibility [2–5]. Many of KOs [20]. However, cell density measurements are un- these genes cluster upon proteins relating to synaptic sig- likely to reveal the true extent of changes within the autis- nalling [6]. A family of presynaptic proteins garnering re- tic brain. Within the current study, we have addressed this cent interest have been the neurexins (NRXNs). NRXNs by developing a dual imaging approach (DTI and CLAR- are encoded by three genes (NRXN1, NRXN2, NRXN3; ITY) that quantifies the alignment and density of white note that CNTNAP1 and CNTNAP2 are sometimes re- matter, applied here to brain regions known to support so- ferred to as NRXN4), of which two major isoforms exist: cial behaviour in a mouse model of autism. the longer α proteins with six laminin/neurexin/sex hor- Diffusion tensor MRI (or DTI) is based upon the mone (LNS) binding domains, and the shorter β proteins movement of water molecules, a measure that is termed with one LNS binding domain [7, 8]. fractional anisotropy (FA). Apparent diffusion coefficient Mutations within all three NRXN genes have been linked (ADC) is similar to FA but quantifies diffusion restric- to autism [6]. Heterozygous deletions within NRXN2 have tion as opposed to the spatial symmetry of diffusion. been identified in a number of individuals with autistic phe- This approach has been used to explore neuropatho- notypes. These include an autistic boy and his father (who logical markers in autistic patients; alterations in myelin- had severe language delay but not autism) who both had a ation, axonal abundance, size and orientation all modify frameshift mutation within exon 12 of NRXN2 [9]; a FA and ADC values [21–23]. Using the preferred direc- 570-kb de novo deletion of 24 genes at chromosome tion of the diffusion of tensors between brain regions 11q13.1, including NRXN2, in a 21-year-old man displaying can be used to explore their potential connection. Quan- a clinical phenotype including autistic traits [10]; a 1.6-Mb tification of those computed streamlines by FA and axial deletion at chromosome region 11q12.3–11q13.1, including and/or radial diffusion can indicate impairments in re- NRXN2, in a 23-year-old man with intellectual disability gional structural connectivity. Since aberrant brain con- and behavioural problems [11]; a de novo frameshift muta- nectivity is likely a core feature of autism [24], we tion identified in a Chinese man with autism spectrum dis- reasoned that the candidate method for probing the aut- order (ASD) [12], a 921-kb microdeletion at 11q13 in a istic brain should combine tractographic techniques. Ac- 2-year-old boy who had language and developmental delay cordingly, here, we combined high-resolution imaging of (although did not meet the autism diagnosis criteria) [13] labelled neuronal tracts in brains rendered transparent and a paternally inherited microRNA miR-873-5p variant by CLARITY with DTI. in an ASD individual which altered binding affinity for sev- CLARITY is a recent development that renders tissue eral risk genes including NRXN2 and CNTNAP2 (NRXN4) optically transparent and macromolecule permeable [14]. Furthermore, recently, two large-scale reports have [25]. This permits antibody staining and imaging of identified NRXN2 with ASD risk. A study of 529 ASD pa- much larger tissue volumes than possible under trad- tients and 1923 controls in a Chinese population identified itional immunofluorescence techniques. By examining two NRXN2 variants which significantly increase ASD risk fibre orientation without sectioning-related artefacts and [15]. The second study employed machine learning ap- biases, axonal staining in cleared tissue affords a deeper proaches across 5000 ASD families to rank the importance understanding of the microstructure and structural con- of ASD candidate genes and ranks NRXN2 in the top ~ nectivity of a brain region. 0.5% of genes, i.e. 113th [16]. For comparison, NRXN1,for Given the social impairments found within Nrxn2α which the evidence base for its links to ASD is broader and mice, we sought to examine those brain regions most stronger, ranks 45, and CNTNAP2 ranks 211th [16]. Con- closely linked with social behaviour (see Additional file 1: sistent with these association studies, we and others have Supplemental materials). Briefly, we identified four re- previously found that homozygous or heterozygous gions of interest (ROIs): the amygdala and three brain deletion of Nrxn2α induces impairment in social ap- regions strongly and directly connected to the amygdala; proach and social recognition [17–19]. In summary, the hippocampus, orbitofrontal cortex (OFC), and anter- although mutations within NRXN2 are rare, under- ior cingulate cortex (ACC). As predicted, structural con- standing how they may drive social, ASD-relevant be- nectivity was abnormal in Nrxn2α mice. havioural changes is important. One important goal is to help elucidate how apparently convergent patho- Methods physiology in ASD emerges despite marked genetic Ethics heterogeneity [5]; mapping brain alterations driven by All procedures were approved by the University of Leeds different single genes is thus a crucial task. and Durham University Animal Ethical and Welfare Pervolaraki et al. Molecular Autism (2019) 10:8 Page 3 of 13

Review Boards and were performed under UK Home Of- gradient vector on the x, y and z orientations. Unwanted fice Project and Personal Licenses in accordance with background, setting a threshold, smoothing of the data the Animals (Scientific Procedures) Act 1986. and definition of tissue boundaries were performed prior to the reconstruction of the final 3D image. DTI analysis Animals parameters were calculated as previously described [27]. Full details of the animals, their background, genotyping The ex vivo mouse brain 3D diffusion-weighted images and housing can be found elsewhere [17]. In brief, male were reconstructed from the Bruker binary file using DSI B6;129-Nrxn3tm1Sud/Nrxn1tm1Sud/Nrxn2tm1Sud/J Studio (http://dsi-studio.labsolver.org)[28]. Direction mice (JAX #006377) were purchased from the Jackson Encoded Colour Map (DEC) images were generated by Laboratory and outbred once to the C57BL/6NCrl strain combining the information from the primary eigenvectors, (Charles River, Margate, UK) to obtain mice that were diffusion images and the FA. Images of the primary vectors individually Nrxn2α KO heterozygotes. Subsequently, and their orientation were reconstructed and superimposed HET knockout males were bred with HET females on the corresponding FA images to guide the segmentation (cousin mating). of discrete anatomical locations according to the brain atlas (Fig. 1b–d). Region of interest definition was performed by Experimental animals author EP and corroborated independently by JD, with re- Six adult wild-type males (Charles River, Margate, UK) gion area compared between the experimenters (data not and six age-matched littermate Nrxn2α KO homozy- shown). For whole brain region analysis, we used a similar gotes (71 days ± 6 days old (SEM)) were perfused fixed approach, except regions were segmented for every other with 4% paraformaldehyde (PFA) in 0.1 M phosphate slice in the anterior to posterior extent (Fig. 1a–d; Add- buffer saline (PBS) and the brains extracted. The brains itional file 1:FigureS1)[29]. The DSI Studio DTI recon- were immersed in 4% PFA/0.1 M PBS for a minimum of struction characterises the major diffusion direction of the 48 h prior to imaging. Mouse weights were not specific- fibrewithinthebrain[30, 31]. Extraction of FA (calculated ally taken prior to perfusion. However, in a separate co- [26]) and ADC was performed within selected segmented hort, wild-type and Nrxn2α KO homozygotes did not brain areas for every 3D-reconstructed mouse brain. significantly differ in body mass (wild type, n = 15, 30.9 ± 4.1 g; Nrxn2 KO, n = 10, 28.6 ± 4.3 g, t test p = 0.167). Regions of interest (ROIs) We did not specifically time perfusions, although as a Our DTI approach was to undertake an a posteriori ana- matter of process, each mouse was perfused with ~ 60 lysis of neural organisation in regions of interest (ROIs) ml of fixative. We cannot rule out that variance in perfu- identified by previous literature as socially relevant. sion timings may have influenced the results, which is a Given the social impairments found within Nrxn2α limitation of the current study. During imaging, the mice, for the current study, we identified the brain re- samples were placed in custom-built MR-compatible gions of interest (ROIs) most closely linked with social tubes containing Fomblin Y (Sigma, Poole, Dorset, UK). behaviour, using previously published reports of brain Due to the relatively low variance and owing to the region involvement in social behaviour. Quantification complexity and methodological nature in our experi- of c-Fos immunoreactivity has highlighted the import- mental approach, we achieved significance by groups of ance of several amygdala nuclei (including the basolat- 6 (power provided in the ‘Results’ section). No data was eral) following social exposure [32], but also the anterior excluded from the study. Sample randomisation was per- cingulate cortex (ACC), prefrontal cortex and the hippo- formed by JD, with experimenters (EP and ALT) blinded campus [33]. Lesions to the primate amygdala alter so- to genotype. cial interactions [34, 35], and amygdala neurons in primates including humans increase firing rates during Data acquisition social scenarios [36–38]. Consistent with these animal Image acquisition has been described elsewhere [26]. studies, amygdala damage in humans [39] and amygdala Each brain was 3D imaged using the protocol TE 35 ms, dysfunction in ASD patients [40] impair social re- TR 700 ms and 10 signal averages. The field of view was sponses. Other socially important brain regions have also set at 128 × 128 × 128, with a cubic resolution of been proposed. Notably, several studies have implicated 100 μm/pixel and a b value of 1200 s/mm2. Further de- the rodent hippocampus in social behaviour, including tails can be found in Additional file 1: Supplemental social memory and sociability [41–43]. For instance, materials. intrahippocampal administration of neurolide-2, which interacts with α-neurexin, specifically impairs sociability, Image processing but not anxiety and spatial learning in rats [44]. These Parsing of the raw data was semi-automated using DSI findings are consistent with reports of social deficits in Studio, in order to obtain b values for every normalised humans with hippocampal damage [45] and hippocampal Pervolaraki et al. Molecular Autism (2019) 10:8 Page 4 of 13

Fig. 1 Quantification of CLARITY imaging. a Sections of DTI-scanned brain were segmented at different bregma levels for (i) the orbitofrontal cortex, (ii) the anterior hippocampus and amygdala, (iii) the mid hippocampus and posterior amygdala and (iv) the posterior hippocampus. b–d DTI-scanned brains were computed for tracts. Tissue from wild-type and Nrxn2α KO mice were cleared and stained for neurofilament and DAPI (e). f Automated MATLAB scripts were used to segment the DAPI (blue) and neurofilament (purple) channels such that cell density and axonal density and orientation could be calculated. g is representative of a CLARITY-derived 3D stacked image of a DAPI and neurofilament of a region of interest, with h being the corresponding segmented image. Scale bar, 100 μm abnormalities in ASD [46, 47]. Finally, several studies link in hydrogel solution at 4 °C prior to polymerisation at the frontal cortex, particularly the orbitofrontal cortex, 37 °C for 3.5 h. The tissue was cut into 1.5-mm coronal which is strongly anatomically connected with the amyg- sections using a custom 3D-printed brain-slicing matrix dala [48], to social processing [49, 50], consistent with find- based on MRI scans of an adult C57BL/6 mouse brain ings of abnormalities in orbitofrontal cortex in ASD [48, [52] and incubated in clearing buffer for 24 days at 37 °C 51]. Control regions of the primary motor cortex (M1), pri- with shaking. The cleared tissue was then washed in mary sensory cortex (S1) and the barrel field were chosen PBSTN3 (0.1% TritonX-100 and 1.5 mM sodium azide in for CLARITY (Additional file 1:FigureS7N–O). PBS) for 24 h at room temperature and incubated in primary antibody solution (neurofilament (Aves NF-H) Clarity 1:100 in PBSTN3) at 37 °C with shaking for 13 days. Following MR imaging, the brains were washed in PBS Samples were washed and then incubated in secondary to remove all Fomblin Y and then incubated for 7 days antibody (AlexaFluor 488 goat anti-chicken IgY) as per Pervolaraki et al. Molecular Autism (2019) 10:8 Page 5 of 13

the primary. Sections were washed again and incubated axis), which is sensitive to changes such as altered align- in 3.6 μM DAPI (4′,6-diamidino-2-phenylindole) ment. The amygdala is critically important for social be- followed by 85% glycerol in PBS for refractive index haviours. To assess whether amygdalar alterations might matching. account for social impairments in Nrxn2α KO mice, we Cleared samples were imaged using a Zeiss 7MP mul- segmented the whole amygdala structure and the baso- tiphoton microscope at 770 nm using a 20 × objective lateral nuclei along the anterior-posterior axis. lens (W Plan-Apochromat, NA 1.0, WD 1.7 mm). Im- The amygdala showed a significant increase in FA in ages (512 × 512 × 512 voxels or 265 × 265 × 265 μm with Nrxn2α KO mice (Fig. 2a) (genotype (F(1, 10) = 11.15, p = an isotropic resolution of 520 nm) were acquired in 0.022, power = 85.2%)). There was a FA reduction ob- ACC, basolateral (BLA) and basomedial amygdala and served specifically in the BLA, a region strongly associ- OFC in both hemispheres. DAPI and neurofilament sig- ated with social behaviours (Fig. 2b; genotype (F(1, 10) = nal was segmented into cell nuclei and axons, and the 6.31, p = 0.049)). ADC was not significantly altered in resulting binary images were used to generate values for the whole amygdala or BLA (Fig. 2c, d; all genotype F(1, cell density, axonal density and axonal alignment. 10) < 1). Full CLARITY methodological details are available We conducted the same analysis for the two prefrontal within Additional file 1: Supplemental materials. regions implicated in social behaviour and autism: the OFC and ACC. The pattern of results was similar for Data availability both regions: FA was significantly altered, while ADC Codes to analyse CLARITY datasets are made available was unaffected (Fig. 3a, b) and the ACC (Fig. 3e, f). FA by author LCA by email request to either JD or LCA, for the OFC was significantly increased (genotype (F(1, subject to reference to the current paper. The datasets 10) = 16.14, p = 0.009, power = 95.0%)), but ADC was used and/or analysed during the current study are avail- similar between the genotypes (genotype (F(1, 10) = 1.43, able from the corresponding author on reasonable p = 0.11)). The ACC also had significantly increased FA request. (t(10) = 2.55, p = 0.03, power = 71.0%), but ADC was unaltered (t(10) = 0.51, p = 0.618). Data analysis We sought to examine whether changes in the amygdala, All data are expressed as mean ± standard error of the OFC or ACC FA and ADC were driven by diffusion in the mean (SEM). To assess the variance between genotypes primary axis (λ1) or the radial orientations (λ2 and λ3)by within a single brain structure across hemispheres (given characterisation of AD (primary) and RD (radial). Within the importance of hemispheric differences in ASD [53]), the amygdala, neither AD nor RD was significantly altered data was analysed by within-subject repeated measures in Nrxn2α KO mice (Fig. 2e; AD genotype: F(1, 10) = 3.06, two-way ANOVAs, with Sidak multiple corrections p = 0.111, Fig. 2f; RD genotype: F(1, 10) =2.47, p = 0.147). employed on post hoc testing, or unpaired t tests. To Within the OFC (Fig. 3c, d), AD was significantly increased correct for multiple comparisons, we employed the (genotype (F(1, 10) =6.71, p =0.032, power=64.7%)), Benjamini-Hochberg procedure (corrected p values whereas RD was significantly decreased (genotype (F(1, stated). Non-significant statistical results, particularly 10) = 10.07, p = 0.025, power = 81.5%)), suggesting that both hemisphere comparisons, can be found in Add- along-tract diffusion and tract branching were affected. itional file 1: Supplemental materials. Statistical testing However, in the ACC (Fig. 3g–h), only AD was significantly and graphs were made using GraphPad Prism version 6 increased (t(10) =3.89,p = 0.019, power = 96.9%), with no al- and SPSS v22. teration in RD (t(10) = 1.35, p = 0.10). Increased AD and decreased RD is thought to reflect changes in axonal density Results or orientation [54]. Nrxn2α deletion disrupts DTI measures of microstructure in social brain regions DTI reveals altered hippocampal microstructure in Nrxn2α To assess whether Nrxn2α deletion alters gross morph- KO mice ology, we quantified whole brain volume using DTI. We The hippocampus has recently been associated with so- found total brain volume for wild types and Nrxn2α cial motivation and social recognition. Since the specific KOs was similar (456.0 ± 14.76 vs. 466.2 ± 11.0 mm3 (re- contributions of the dorsal and ventral hippocampal spectively); t(10) = 0.55, p = 0.59). Thus, Nrxn2α deletion poles remain unclear, we segmented the whole hippo- does not change the total brain size. campus into anterior (Bregma − 1.06 to − 2.46 mm) (in- To quantitatively measure DTI, we examined FA and corporating dorsal) and posterior (Bregma − 2.54 to − ADC. FA analyses changes in the linear orientation (i.e. 3.16 mm) (incorporating ventral regions) levels. along an axonal tract), whereas ADC (mean diffusivity) FA values in the anterior and posterior hippocampus averages diffusion in all directions (i.e. the X-, Y- and Z- were significantly increased (Additional file 1: Figure Pervolaraki et al. Molecular Autism (2019) 10:8 Page 6 of 13

Fig. 2 Deletion of Nrxn2α increases amygdala fractional anisotropy (FA) but not apparent diffusion coefficient (ADC). DTI images of the amygdala was segmented at two regions; the whole amygdala in the anterior to posterior extent or the basolateral amygdala (BLA) centred at bregma − 1.94 mm. FA of the whole amygdala structure was significantly increased (a) but was decreased in the BLA (b). However, ADC was similar between the genotypes (c and d). Axial (AD) (e) and radial diffusivity (RD) (f) was unaltered in the amygdala. **p < 0.01, *p < 0.05. Error bars represent s.e.m. Wild type n =6,Nrxn2α KO n =6

S4A and E; see figure legend for statistics). However, In summary, the microstructural measures most al- ADC was unaltered for the anterior and posterior hippo- tered by Nrxn2α deletion were increases in FA, AD and campus (Additional file 1: Figure S4B and F). AD was RD, including in the hippocampus, in line with recent significantly increased in both the anterior and posterior work suggesting a role for ventral hippocampus in social hippocampal regions (Additional file 1: Figure S4C and memory [43]. G). RD was also significantly decreased in the anterior and posterior hippocampus in Nrxn2α KO mice (Add- itional file 1: Figure S4D and H). DTI tractography reveals Nrxn2α deletion affects Lastly, given DTI is most commonly associated with structural connectivity between the amygdala and analysis of white matter tracts, we also quantified the orbitofrontal cortex corpus callosum. Changes within the corpus callosum The amygdala is strongly and bidirectionally connected have repeatedly been highlighted in autism [55, 56], in- to both the hippocampus [59] and the OFC [60]. As all cluding mouse models of autism [57, 58]. Here, we three regions are themselves important for social behav- found significantly increased FA and reduced ADC in iour, and autism is thought to be, at least in part, related Nrxn2α KO mice, which were driven by a significant re- to abnormal structural connectivity [24], we performed duction in RD (Additional file 1: Figure S6). tractography analysis between the amygdala (and Pervolaraki et al. Molecular Autism (2019) 10:8 Page 7 of 13

Fig. 3 Nrxn2α KO mice have increased fractional anisotropy (FA) and axial (AD) and radial diffusivity (RD) in the orbitofrontal cortex (OFC) and the anterior cingulate cortex (ACC). FA was significantly different between wild-type and Nrxn2α KO mice for FA in the OFC (a) and ACC (e), but ADC was not significantly changed in Nrxn2α KO mice in both prefrontal regions (b and f). The OFC has significantly increased AD and RD (c and d), whereas only AD was increased in the ACC (g–h). **p < 0.01, *p < 0.05. Error bars represent s.e.m. Wild type n =6,Nrxn2α KO n =6 specifically the BLA) and the hippocampus, and between did not observe differences in RD in the tracts connect- the amygdala and the OFC. ing the amygdala with the hippocampus (see Add- From the anterior amygdala, we examined the diffusiv- itional file 1: Table S1 for non-significant statistics). ity (AD and RD) of connections to the anterior and pos- Although AD between the anterior amygdala and anter- terior hippocampus (Additional file 1: Figure S6). We ior hippocampus did not differ by genotype, there was a Pervolaraki et al. Molecular Autism (2019) 10:8 Page 8 of 13

Fig. 4 Tractographic analysis of amygdala-hippocampus and amygdala-orbitofrontal cortex (OFC) connectivity. Amygdala-hippocampal connections are characterised by greater right hemisphere axial diffusivity (AD) in Nrxn2α KO mice (a) but not radial diffusivity (RD) (b). Specific to the BLA, connections to the anterior hippocampus (c) and posterior hippocampus (d) have greater right hemisphere AD. Although the amygdala- OFC connection was similar between the genotypes for AD (e), Nrxn2α KO mice had significantly increased RD (f). *p < 0.05, ***p < 0.001. Error bars represent s.e.m. Wild type n = 6, Nrxn2α KO n =6 significant interaction between the genotype and hemi- Finally, we tested connections between the amygdala sphere (genotype × hemisphere (F(1, 10) = 12.12, p = and the OFC. For AD, wild-type and Nrxn2α KO mice 0.023, power = 88.0%; Fig. 4a); post hoc analysis shows did not differ by genotype (Fig. 4e; genotype (F(1, 10) = this was driven by larger right-vs-left hemisphere AD 2.85, p = 0.09), hemisphere (F(1, 10) = 6.38, p = 0.052). RD values within the Nrxn2α KOs only (p = 0.012). This dif- was strikingly higher in Nrxn2α KO mice (Fig. 4f; geno- ference could be driven by the BLA; there was increased type (F(1, 10) = 26.06, p = 0.023, power = 99.5%)), indica- AD in both the BLA/anterior hippocampus tracts (geno- tive of a change in demyelination, axonal density or type × hemisphere (F(1, 10) = 10.53, p = 0.032, power = orientation [54]. 83.2%) and the BLA/posterior hippocampus tracts (genotype × hemisphere (F(1, 10) = 12.97, p = 0.020, CLARITY reveals fibre disruption in Nrxn2α KO mice in the power = 90%), which again was related to larger amygdala, orbitofrontal cortex and anterior cingulate right-vs-left hemisphere values in the Nrxn2α KOs cortex (BLA/anterior hippocampus p = 0.004 and BLA/poster- To further explore the differences as revealed by DTI, ior hippocampus p = 0.001, (Fig. 4c–d)) but not the wild we performed CLARITY on the same brain tissue used type (anterior p = 0.87; posterior p = 1.00). These results in DTI and stained with neurofilament and DAPI to indicate that there are differences for the structural con- label axons and cell bodies, respectively. We were then nectivity of the amygdala with the hippocampus within able to derive both the axonal alignment (as in, the geo- theleftandrighthemisphereinNrxn2α KO mice, metric alignment of axons (from linear alignment to ran- with increased axial diffusivity in the right hemi- dom) within 3D space (see Additional file 1: Figure S2)) sphere. This finding is particularly interesting, as and density of the stained fibres, in addition to the cell hemispheric differences in functional connectivity, density. particularly affecting connections from the right The pattern of results was broadly similar for both the amygdala, have been found in children with ASD prefrontal cortical ROIs. That is, first, axonal alignment [61, 62]. was increased in Nrxn2α KO mice in the ACC (Fig. 5d; Pervolaraki et al. Molecular Autism (2019) 10:8 Page 9 of 13

Fig. 5 CLARITY reveals differences in axonal alignment and fibre density in Nrxn2α KO mice. a–c Representative images of the CLARITY-treated brain, with ROI defined for the anterior cingulate cortex (ACC), orbitofrontal cortex (OFC), basomedial amygdala (BMA) and basolateral amygdala (BLA). For the ACC, the axonal alignment (d) and axon density (e) were significantly altered in KO mice, but cell density was unaltered (f). Within the medial OFC, only axonal alignment was significantly altered in KOs (g), with axon density (h) and cell density (i) being similar. For the BMA, both the axonal alignment (j) and axon density (k) were significantly increased, while cell density was unaltered (l). *p < 0.05, **p < 0.01. Error bars represent s.e.m. Wild type n =6,Nrxn2α KO n =6

genotype (F(1, 10) = 16.06, p = 0.011, power = 94.9%) but differences for axonal alignment or fibre density in the not the OFC (Fig. 5g; genotype (F(1, 10) = 5.56, p = BLA (see Additional file 1: Figure S7A–C), whereas 0.059). Second, this could not be explained by a differ- axonal alignment (Fig. 5j; genotype F(1, 10) = 7.70, p = ence in cell density, since that was similar between the 0.045, power = 70.6%) but not axonal density (Fig. 5k; KO and wild-type mice in both the ACC (Fig. 5f; geno- genotype (F(1, 10) = 6.10, p = 0.054) was increased in type (F(1, 10) < 1), hemisphere (F(1, 10) = 1.73, p = 0.11) Nrxn2α KO mice in the basomedial nuclei, while cell and the OFC (Fig. 5h; genotype (F(1, 10) = 3.09, p = 0.08). density was unaffected (Fig. 5l; genotype (F(1, 10) < 1). Al- An increase in axonal density in Nrxn2α KO mice was terations in axonal alignment and density as directly re- reliable in the ACC (Fig. 5e; genotype (F(1, 10) = 14.64, vealed by CLARITY could explain the increases in p = 0.014, power = 93.0%), but not in the OFC (Fig. 5h; diffusivity and RD in the prefrontal regions, as measured genotype (F(1, 10) = 3.09, p = 0.083). by DTI. We further examined two regions of the anterior To test the specificity of these alterations, we exam- amygdala, the BLA and basomedial (BMA) nuclei, where ined three further brain regions: the primary motor cor- altered social cellular responses have been reported in tex (M1; Additional file 1: Figure S7D–F); the primary human autism [38]. We did not observe any significant somatosensory cortex (S1; Additional file 1: Figure S7H– Pervolaraki et al. Molecular Autism (2019) 10:8 Page 10 of 13

J); and the barrel field (BF; Additional file 1: Figure explained by altered axonal patterning (i.e. CLARITY). S7K–M). Interestingly, although there were differences Others have explored the biological mechanisms linking between the hemispheres, there were no statistical differ- structural connectivity to altered functional connectivity. ences between the genotypes or genotype × hemisphere Zhan et al. found that deletion of the chemokine recep- interactions for any measure (Additional file 1: Table tor Cx3cr1 resulted in impaired synaptic pruning of S2), suggesting some specificity of the alterations in long-range connections during development, which social-relevant brain regions in Nrxn2α KO mice. manifested as impaired social behaviour caused by de- In summary, in both the prefrontal ROIs, namely the creased frontal functional connectivity, reduced synaptic OFC and the ACC, DTI showed that ADC and RD were multiplicity and weakened coherence of local field po- increased in Nrxn2α KO mice, likely related to comple- tentials [73]. Thus, it is possible that impairments in mentary analysis from CLARITY showing that axonal neuronal structural maturation can generate functional alignment was altered in Nrxn2α KO mice in both pre- connectivity deficits that encapsulate core autism frontal ROIs. phenotypes. Our findings corroborate these quantifications of clin- Discussion ical autism but highlight the question of what do the dif- Interestingly, the single-gene deletion of Nrxn2α cap- ferent measures of ADC, FA, AD and RD represent? tures several key aspects of human ASD. In terms of be- Importantly, we observed these microstructural changes haviour, three studies have now found social deficits in various socially relevant brain regions against a back- associated with Nrxn2α KO [17–19]; in terms of brain ground of unchanged cell density in all our study’s ROIs. structure, as reported here (summarised below), the Unexpectedly, this highlights the power of our new ap- Nrxn2α KO mouse model shows altered microstructure proach. Dudanova et al. concluded from measures of cell and structural connectivity patterns in socially relevant counting and cortical cell layer thickness that NRXN2 brain regions reminiscent of changes in ASD. played little role in normal brain development [20]. In- A DTI approach has been used for some time to ex- deed, in earlier studies, it was suggested that deletion of plore neuropathological markers in autistic patients; al- all Nrxns was unlikely to affect synaptic development terations in myelination, axonal abundance, size and but instead disrupts synaptic function [74]. We propose orientation all modify FA and ADC values [21, 63], spe- that measures such as two-dimensional cell counting cifically by reducing amygdala FA [23, 63], and have may be underestimating the impact of genetic mutations been used as a quantitative measure of changes to brain upon normal development. By staining cleared brain tis- white matter integrity [23, 24]. However, several studies sue with a nuclear marker and performing automated have noted increases in FA in ASD patients (see Table 1 three-dimensional cell counting, we found no effect of of [64]). Furthermore, both increased RD of various Nrxn2α deletion on cell density in any region of interest white matter tracts [65, 66] and increased whole-brain examined. But this belies the clear effects upon micro- AD [66] have been observed in ASD. The Nrxn2α KO structure integrity across multiple regions as measured mouse reproduces some of these specific changes, in- by both DTI and CLARITY and its specificity; only the cluding altered FA and increases in ADC, AD and RD. socially relevant brain regions we tested were disrupted, Whole brain increases in ADC, AD and RD (but not FA) and not primary sensory or motor regions. Future stud- have been reported in ASD children, as have increases in ies will benefit from employing more sensitive measures ADC and RD in frontal cortex tracts [66]. FA has been of brain structural connectivity to determine the rele- noted as reduced in the amygdala in ASD children and vance of genetic mutations in development. adolescents [67], and right-sided lateralisation of abnor- FA and ADC can be influenced by changes in axonal mal amygdala/hippocampus-related connections, as seen density and alignment (e.g. by myelination, demyelin- in our Nrxn2α KO mouse, has been noted in ation, axonal damage, loss of white matter coherence high-functioning adolescents/adults with autism [68]. [75]). It is likely that the axonal alignment metric used While the current study specifically explores structural to quantify CLARITY more closely reflects the ADC connectivity, it is difficult to extrapolate as to what these measure of DTI, given that ADC (or mean diffusivity) structural changes mean for functional connectivity in equally weighs diffusion across all eigenvectors and does the Nrxn2α KO mouse. Hyper and hypo connectivity not bias the primary eigenvector as FA does. Thus, it is theories of autism have remained contentious and vary likely that alterations in the properties of axons in in humans by cohort studied (e.g. by age of participant) Nrxn2α KO mice are driving these changes in FA and [69]. Further, in studies that have combined resting-state ADC. Given we see differences in RD, thought to reflect λ functional MRI (rsfMRI) and DTI, functional and struc- tract branching and myelination (as it measures 2 and – tural connectivity do not always overlap [70 72]. Our λ3), it is possible that the orientation in the perpendicu- current data suggests that DTI differences can be lar not parallel orientation of fibres is mostly affected. Pervolaraki et al. Molecular Autism (2019) 10:8 Page 11 of 13

Given the differences in the amygdala, OFC and ACC, it study of this nature. Despite this, as the adoption of the is possible that even though neuronal densities are simi- CLARITY technique increases, we hope that the use of lar in the Nrxn2α KO brain, it is the connections be- DTI and CLARITY to study structural connectivity tween neurones and brain regions that are perturbed. across spatial scales will become a commonplace. This would be consistent with the idea that structural As yet, no one DTI protocol has emerged as the stand- connectivity disruption may represent a core feature of ard for in vivo or ex vivo imaging. Indeed, there has autism [76]. A broader question is how does the loss of been debate regarding the best number of diffusion gra- Nrxn2α account for changes in axonal organisation? Ul- dients to use, among other parameters [79]. Undoubt- timately, this question requires further studies. Others edly, more directions that what we used here would have shown that in Nrxn2α KO mice, excitatory trans- facilitate better interpretations, this is a limitation of the mitter release is reduced, as is short-term plasticity [18]. current work. Despite this, the major purpose of the Reduced glutamatergic release, even at a relatively long current paper is to develop a new generation of CLAR- range to the synapse, can change the complexity of den- ITY analysis. We hope that future studies will refine on dritic arbours [77]. As this is a gene deletion model, it is both DTI and CLARITY parameters to maximise ana- conceivable that altered glutamatergic signalling during lysis methodology. A further potential limitation of the early development impairs appropriate synapse matur- current study is that groups of six animals may be ation, leading to the structural changes we see herein. underpowered. We argue for our approach here as fol- Further, how or whether these structural changes fully lows. First, low variance in the datasets permits smaller explain the social impairments of Nrxn2α KO mice group sizes. Second, for most of our significant results, would require new studies. Conceivably, inducible the observed power was more than 80%. Third, given knockdown of Nrxn2 (by inducible knockout, siRNA, the technical complexity of this approach, particularly in optogenetics etc.) within a specific brain region would its early adoption and refinement stages, large sample provide evidence that social abnormalities are being throughput of multiple brain regions is challenging. driven by Nrxn2 loss. However, developmentally In summary, our combined use of DTI and CLARITY dependent altered structural connectivity would be has revealed changes in microstructure and structural harder to definitively manipulate to explain changes in connectivity of socially relevant brain regions in Nrxn2α social behaviours. KO mice that may underlie their deficits in social behav- Here we have developed a new application of CLAR- iour. It is hard to conceive how these changes could ITY to quantitatively investigate disease models by com- have been observed using classical experimental ap- bining DTI with high-resolution 3D imaging and proaches. We envisage this approach will deliver a new automated analysis of axonal fibres in a within-subject level of detail in structural connectivity approaches to study. Inevitably, there are some technical limitations understanding autism. that will require future refinement as this technology matures. Additional file First, while we used CLARITY and immunolabelling to identify axons, we cannot know whether axon-related Additional file 1: Supplemental materials and methods changes alone reflect all the changes we observed for (DOCX 6377 kb) our DTI measures. Second, while we can segment entire brain regions for DTI analysis, it was not practical to Abbreviations ACC: Anterior cingulate cortex; AD: Axial diffusivity; ADC: Apparent diffusion image larger brain areas at the necessary resolution for coefficient; ASD: Autism spectrum disorder; BLA: Basolateral amygdala; CLARITY. While it is theoretically possible that we may CLARITY: Optically cleared brain tissue; DTI: Diffusion tensor imaging; bias sampling of each brain region by picking ROIs for FA: Fractional anisotropy; Nrxn2: Neurexin II; OFC: Orbitofrontal cortex; RD: Radial diffusivity; ROI: Region of interest multiphoton imaging, this was done using atlas-defined coordinates and by an experimenter blind to the DTI re- Acknowledgements sults, so minimising any bias. However, within the Not applicable. current study, we were only able to apply the CLARITY Funding approach to the amygdala, OFC and ACC. It was not This work was supported by the Guy’s and St. Thomas’ Charity Prize PhD practical to apply this methodology to the hippocampus, scholarship to ALT, a Medical Research Council (UK) grant (G0900625) to SJC due to its extremely heterogeneous structure. The small and RJR, a University of Leeds Wellcome Trust ISSF (UK) Fellowship, a Royal Society (UK) grant (RG130316), an Alzheimer’s Society Fellowship (AS-JF-15- cubic ROIs could not be reproducibly positioned, and 008) to JD, a British Pharmacological Society (UK) grant to JD and CL, a larger ROIs to average across larger areas of the hippo- BBSRC grant to LCA (BB/P000479/1) and a BBSRC grant to CL (BB/M008975/ campus were not possible. Although imaging of fibre 1). We acknowledge financial support from the Innovative Medicines Initiative Joint Undertaking under grant agreement no. 115300, resources of tracts in large volumes of cleared tissue is possible [78], which are composed of financial contribution from the European Union’s fluorescent labelling limitations make it impractical for a Seventh Framework Programme (FP7/2007–2013) and EFPIA companies’ in Pervolaraki et al. Molecular Autism (2019) 10:8 Page 12 of 13

kind contribution, the Mortimer D Sackler Foundation and the Sackler 9. Gauthier J, Siddiqui TJ, Huashan P, Yokomaku D, Hamdan FF, Champagne N, Institute for Translational Neurodevelopment (ALT and LCA). Some analysis et al. Truncating mutations in NRXN2 and NRXN1 in autism spectrum scripts were provided to ALT at the Computational Image Analysis in Cellular disorders and schizophrenia. Hum Genet. 2011;130(4):563–73. and Developmental Biology course at the Marine Biological Laboratory 10. Mohrmann I, Gillessen-Kaesbach G, Siebert R, Caliebe A, Hellenbroich Y. A (Woods Hole, MA, USA), funded by National Institutes of Health (R25 de novo 0.57 Mb microdeletion in chromosome 11q13.1 in a patient with GM103792-01). speech problems, autistic traits, dysmorphic features and multiple endocrine neoplasia type 1. Eur J Med Genet. 2011;54(4):e461–4. Availability of data and materials 11. Boyle MI, Jespersgaard C, Nazaryan L, Ravn K, Brondum-Nielsen K, Bisgaard The codes used to quantify the CLARITY datasets are made available by AM, et al. Deletion of 11q12.3-11q13.1 in a patient with intellectual disability author LCA by email request to authors LCA or JD, subject to reference to and childhood facial features resembling Cornelia de Lange syndrome. the current paper. The datasets used and/or analysed during the current Gene. 2015;572(1):130–4. study are available from the corresponding author on reasonable request. 12. Li J, Wang L, Gou H, Shi L, Zhang K, Tang M, et al. Targeted sequencing and functional analysis reveal brain-size-related genes and their networks in Authors’ contributions autism spectrum disorders. Mol Psychiatry. 2017;22:1282–90. EP, ALT, LCA and JD conceived the study. EP and ALT performed the 13. Yuan H, Li X, Wang Q, Yang W, Song J, Hu X, et al. A de novo 921Kb experiments. EP, ALT, LCA and JD analysed the data. SJC, RJR, LCA and JD microdeletion at 11q13.1 including neurexin 2 in a boy with developmental funded the study. All authors contributed to writing the paper. All authors delay, deficits in speech and language without autistic behaviors. Eur J Med have read and approved the final manuscript. Genet. 2018;61(10):607–11. 14. Williams SM, An JY, Edson J, Watts M, Murigneux V, Whitehouse AJO, et al. Ethics approval and consent to participate An integrative analysis of non-coding regulatory DNA variations associated All experiments were performed under UK Home Office Project and Personal with autism spectrum disorder. Mol Psychiatry. 2018. https://doi.org/10. Licenses in accordance with the Animals (Scientific Procedures) Act 1986, 1038/s41380-018-0049-x. and with the approval of the University of Leeds and Durham University 15. Wang J, Gong J, Li L, Chen Y, Liu L, Gu H, et al. Neurexin gene family variants – Animal Ethical and Welfare Review Boards. as risk factors for autism spectrum disorder. Autism Res. 2018;11(1):37 43. 16. Duda M, Zhang H, Li HD, Wall DP, Burmeister M, Guan Y. Brain-specific Consent for publication functional relationship networks inform autism spectrum disorder gene Not applicable prediction. Transl Psychiatry. 2018;8(1):56. 17. Dachtler J, Glasper J, Cohen RN, Ivorra JL, Swiffen DJ, Jackson AJ, et al. Deletion of alpha-neurexin II results in autism-related behaviors in mice. Competing interests Transl Psychiatry. 2014;4:e484. The authors declare that they have no competing interests. 18. Born G, Grayton HM, Langhorst H, Dudanova I, Rohlmann A, Woodward BW, et al. Genetic targeting of NRXN2 in mice unveils role in excitatory cortical Publisher’sNote synapse function and social behaviors. Front Synaptic Neurosci. 2015;7:3. Springer Nature remains neutral with regard to jurisdictional claims in 19. Dachtler J, Ivorra JL, Rowland TE, Lever C, Rodgers RJ, Clapcote SJ. Heterozygous published maps and institutional affiliations. deletion of alpha-neurexin I or alpha-neurexin II results in behaviors relevant to autism and schizophrenia. Behav Neurosci. 2015;129(6):765–76. Author details 20. Dudanova I, Tabuchi K, Rohlmann A, Sudhof TC, Missler M. Deletion of 1School of Biomedical Sciences, University of Leeds, Leeds LS2 9JT, UK. alpha-neurexins does not cause a major impairment of axonal pathfinding 2Centre for Developmental Neurobiology, Institute of Psychiatry, Psychology or synapse formation. J Comp Neurol. 2007;502(2):261–74. and Neuroscience, King’s College London, London SE1 1UL, UK. 3MRC Centre 21. Beaulieu C. The basis of anisotropic water diffusion in the nervous system - for Neurodevelopmental Disorders, King’s College London, London SE1 1UL, a technical review. NMR Biomed. 2002;15(7–8):435–55. UK. 4Department of Forensic and Neurodevelopmental Sciences, Institute of 22. Barnea-Goraly N, Kwon H, Menon V, Eliez S, Lotspeich L, Reiss AL. White Psychiatry, Psychology and Neuroscience, King’s College London, London matter structure in autism: preliminary evidence from diffusion tensor SE5 8AF, UK. 5Department of Psychology, Durham University, South Road, imaging. Biol Psychiatry. 2004;55(3):323–6. Durham DH1 3LE, UK. 6Robarts Research Institute, Western University, 23. Noriuchi M, Kikuchi Y, Yoshiura T, Kira R, Shigeto H, Hara T, et al. Altered London, ON N6A 5B7, Canada. 7School of Psychology, University of Leeds, white matter fractional anisotropy and social impairment in children with Leeds LS2 9JT, UK. autism spectrum disorder. Brain Res. 2010;1362:141–9. 24. Belmonte MK, Allen G, Beckel-Mitchener A, Boulanger LM, Carper RA, Webb Received: 15 October 2018 Accepted: 18 February 2019 SJ. Autism and abnormal development of brain connectivity. J Neurosci. 2004;24(42):9228–31. 25. Chung K, Wallace J, Kim SY, Kalyanasundaram S, Andalman AS, Davidson TJ, References et al. Structural and molecular interrogation of intact biological systems. 1. Sandin S, Lichtenstein P, Kuja-Halkola R, Larsson H, Hultman CM, Nature. 2013;497(7449):332–7. Reichenberg A. The familial risk of autism. Jama. 2014;311(17):1770–7. 26. Pervolaraki E, Anderson RA, Benson AP, Hayes-Gill B, Holden AV, Moore BJ, 2. Cardno AG, Gottesman II. Twin studies of schizophrenia: from bow-and- et al. Antenatal architecture and activity of the human heart. Interface focus. arrow concordances to star wars Mx and functional genomics. Am J Med 2013;3(2):20120065. Genet. 2000;97(1):12–7. 27. Pervolaraki E, Dachtler J, Anderson RA, Holden AV. Ventricular myocardium 3. Ronald A, Happe F, Bolton P, Butcher LM, Price TS, Wheelwright S, et al. development and the role of connexins in the human fetal heart. Sci Rep. Genetic heterogeneity between the three components of the autism 2017;7(1):12272. spectrum: a twin study. J Am Acad Child Adolesc Psychiatry. 2006;45(6):691–9. 28. Yeh FC, Verstynen TD, Wang Y, Fernandez-Miranda JC, Tseng WY. 4. Ronemus M, Iossifov I, Levy D, Wigler M. The role of de novo mutations in the Deterministic diffusion fiber tracking improved by quantitative anisotropy. genetics of autism spectrum disorders. Nat Rev Genet. 2014;15(2):133–41. PLoS One. 2013;8(11):e80713. 5. Geschwind DH, State MW. Gene hunting in autism spectrum disorder: on 29. Paxinos G, Franklin K. The mouse brain in stereotaxic coordinates. 3rd ed. the path to precision medicine. Lancet Neurol. 2015;14(11):1109–20. London: Academic Press; 2008. 6. Sahin M, Sur M. Genes, circuits, and precision therapies for autism and 30. Basser PJ, Mattiello J, LeBihan D. Estimation of the effective self-diffusion related neurodevelopmental disorders. Science. 2015;350(6263). https://doi. tensor from the NMR spin echo. J Magn Reson Series B. 1994;103(3):247–54. org/10.1126/science.aab3897. 31. Jiang H, van Zijl PC, Kim J, Pearlson GD, Mori S. DtiStudio: resource program 7. Lise MF, El-Husseini A. The neuroligin and neurexin families: from structure for diffusion tensor computation and fiber bundle tracking. Comput to function at the synapse. Cell Mol Life Sci. 2006;63(16):1833–49. Methods Prog Biomed. 2006;81(2):106–16. 8. Sudhof TC. Synaptic neurexin complexes: a molecular code for the logic of 32. Ferguson JN, Aldag JM, Insel TR, Young LJ. Oxytocin in the medial amygdala is neural circuits. Cell. 2017;171(4):745–69. essential for social recognition in the mouse. J Neurosci. 2001;21(20):8278–85. Pervolaraki et al. Molecular Autism (2019) 10:8 Page 13 of 13

33. Tanimizu T, Kenney JW, Okano E, Kadoma K, Frankland PW, Kida S. Functional 58. Kumar M, Kim S, Pickup S, Chen R, Fairless AH, Ittyerah R, et al. Longitudinal connectivity of multiple brain regions required for the consolidation of social in-vivo diffusion tensor imaging for assessing brain developmental changes recognition memory. J Neurosci. 2017;37(15):4103–16. in BALB/cJ mice, a model of reduced sociability relevant to autism. Brain 34. Emery NJ, Capitanio JP, Mason WA, Machado CJ, Mendoza SP, Amaral DG. Res. 2012;1455:56–67. The effects of bilateral lesions of the amygdala on dyadic social interactions 59. Pitkanen A, Pikkarainen M, Nurminen N, Ylinen A. Reciprocal connections in rhesus monkeys (Macaca mulatta). Behav Neurosci. 2001;115(3):515–44. between the amygdala and the hippocampal formation, perirhinal cortex, 35. Rosvold HE, Mirsky AF, Pribram KH. Influence of amygdalectomy on social and postrhinal cortex in rat. A review. Ann N Y Acad Sci. 2000;911:369–91. behavior in monkeys. J Comp Physiol Psychol. 1954;47(3):173–8. 60. Cavada C, Company T, Tejedor J, Cruz-Rizzolo RJ, Reinoso-Suarez F. The 36. Mosher CP, Zimmerman PE, Gothard KM. Neurons in the monkey amygdala anatomical connections of the macaque monkey orbitofrontal cortex. A detect eye contact during naturalistic social interactions. Curr Biol. 2014; review. Cereb Cortex. 2000;10(3):220–42. 24(20):2459–64. 61. Abrams DA, Lynch CJ, Cheng KM, Phillips J, Supekar K, Ryali S, et al. 37. Rutishauser U, Mamelak AN, Adolphs R. The primate amygdala in social Underconnectivity between voice-selective cortex and reward circuitry in perception - insights from electrophysiological recordings and stimulation. children with autism. Proc Natl Acad Sci U S A. 2013;110(29):12060–5. Trends Neurosci. 2015;38(5):295–306. 62. Murphy ER, Foss-Feig J, Kenworthy L, Gaillard WD, Vaidya CJ. Atypical 38. Rutishauser U, Tudusciuc O, Wang S, Mamelak AN, Ross IB, Adolphs R. functional connectivity of the amygdala in childhood autism spectrum Single-neuron correlates of atypical face processing in autism. Neuron. 2013; disorders during spontaneous attention to eye-gaze. Autism Res Treat. 2012; 80(4):887–99. 2012:652408. 39. Adolphs R, Tranel D, Damasio AR. The human amygdala in social judgment. 63. Sundaram SK, Kumar A, Makki MI, Behen ME, Chugani HT, Chugani DC. Nature. 1998;393(6684):470–4. Diffusion tensor imaging of frontal lobe in autism spectrum disorder. Cereb 40. Baron-Cohen S, Ring HA, Bullmore ET, Wheelwright S, Ashwin C, Williams SC. Cortex. 2008;18(11):2659–65. The amygdala theory of autism. Neurosci Biobehav Rev. 2000;24(3):355–64. 64. Travers BG, Adluru N, Ennis C, do PM T, Destiche D, Doran S, et al. Diffusion tensor 41. Maaswinkel H, Baars AM, Gispen WH, Spruijt BM. Roles of the basolateral imaging in autism spectrum disorder: a review. Autism Res. 2012;5(5):289–313. amygdala and hippocampus in social recognition in rats. Physiol Behav. 65. Ameis SH, Fan J, Rockel C, Voineskos AN, Lobaugh NJ, Soorya L, et al. Impaired 1996;60(1):55–63. structural connectivity of socio-emotional circuits in autism spectrum disorders: 42. Kogan JH, Frankland PW, Silva AJ. Long-term memory underlying a diffusion tensor imaging study. PLoS One. 2011;6(11):e28044. hippocampus-dependent social recognition in mice. Hippocampus. 2000; 66. Shukla DK, Keehn B, Muller RA. Tract-specific analyses of diffusion tensor 10(1):47–56. imaging show widespread white matter compromise in autism spectrum – 43. Okuyama T, Kitamura T, Roy DS, Itohara S, Tonegawa S. Ventral CA1 neurons disorder. J Child Psychol Psychiatry. 2011;52(3):286 95. store social memory. Science. 2016;353(6307):1536–41. 67. Jou RJ, Jackowski AP, Papademetris X, Rajeevan N, Staib LH, Volkmar FR. 44. van der Kooij MA, Fantin M, Kraev I, Korshunova I, Grosse J, Zanoletti O, Diffusion tensor imaging in autism spectrum disorders: preliminary evidence – et al. Impaired hippocampal neuroligin-2 function by chronic stress or of abnormal neural connectivity. Aust N Z J Psychiatry. 2011;45(2):153 62. synthetic peptide treatment is linked to social deficits and increased 68. Conturo TE, Williams DL, Smith CD, Gultepe E, Akbudak E, Minshew NJ. aggression. Neuropsychopharmacology. 2014;39(5):1148–58. Neuronal fiber pathway abnormalities in autism: an initial MRI diffusion 45. Rubin RD, Watson PD, Duff MC, Cohen NJ. The role of the hippocampus in tensor tracking study of hippocampo-fusiform and amygdalo-fusiform – flexible cognition and social behavior. Front Hum Neurosci. 2014;8:742. pathways. J Int Neuropsychol Soc. 2008;14(6):933 46. 46. Nicolson R, DeVito TJ, Vidal CN, Sui Y, Hayashi KM, Drost DJ, et al. Detection 69. Stanfield AC, McIntosh AM, Spencer MD, Philip R, Gaur S, Lawrie SM. Towards a and mapping of hippocampal abnormalities in autism. Psychiatry Res. 2006; neuroanatomy of autism: a systematic review and meta-analysis of structural – 148(1):11–21. magnetic resonance imaging studies. Eur Psychiatry. 2008;23(4):289 99. 47. Schumann CM, Hamstra J, Goodlin-Jones BL, Lotspeich LJ, Kwon H, 70. Vissers ME, Cohen MX, Geurts HM. Brain connectivity and high functioning Buonocore MH, et al. The amygdala is enlarged in children but not autism: a promising path of research that needs refined models, adolescents with autism; the hippocampus is enlarged at all ages. J methodological convergence, and stronger behavioral links. Neurosci – Neurosci. 2004;24(28):6392–401. Biobehav Rev. 2012;36(1):604 25. 48. Bachevalier J, Loveland KA. The orbitofrontal-amygdala circuit and self- 71. Rudie JD, Brown JA, Beck-Pancer D, Hernandez LM, Dennis EL, Thompson PM, et al. Altered functional and structural brain network organization in regulation of social-emotional behavior in autism. Neurosci Biobehav Rev. – 2006;30(1):97–117. autism. NeuroImage Clin. 2012;2:79 94. 72. Uddin LQ, Supekar K, Menon V. Reconceptualizing functional brain 49. Beer JS, John OP, Scabini D, Knight RT. Orbitofrontal cortex and social connectivity in autism from a developmental perspective. Front Hum behavior: integrating self-monitoring and emotion-cognition interactions. Neurosci. 2013;7:458. J Cogn Neurosci. 2006;18(6):871–9. 73. Zhan Y, Paolicelli RC, Sforazzini F, Weinhard L, Bolasco G, Pagani F, et al. 50. Watson KK, Platt ML. Social signals in primate orbitofrontal cortex. Curr Biol. Deficient neuron-microglia signaling results in impaired functional brain 2012;22(23):2268–73. connectivity and social behavior. Nat Neurosci. 2014;17(3):400–6. 51. Girgis RR, Minshew NJ, Melhem NM, Nutche JJ, Keshavan MS, Hardan AY. 74. Missler M, Zhang W, Rohlmann A, Kattenstroth G, Hammer RE, Gottmann K, Volumetric alterations of the orbitofrontal cortex in autism. Prog Neuro- et al. Alpha-neurexins couple Ca2+ channels to synaptic vesicle exocytosis. Psychopharmacol Biol Psychiatry. 2007;31(1):41–5. Nature. 2003;423(6943):939–48. 52. Tyson AL, Hilton ST, Andreae LC. Rapid, simple and inexpensive production 75. Alba-Ferrara LM, de Erausquin GA. What does anisotropy measure? Insights of custom 3D printed equipment for large-volume fluorescence microscopy. from increased and decreased anisotropy in selective fiber tracts in Int J Pharm. 2015;494(2):651–6. schizophrenia. Front Integr Neurosci. 2013;7:9. 53. Herbert MR, Ziegler DA, Deutsch CK, O'Brien LM, Kennedy DN, Filipek PA, 76. Minshew NJ, Williams DL. The new neurobiology of autism: cortex, et al. Brain asymmetries in autism and developmental language disorder: a connectivity, and neuronal organization. Arch Neurol. 2007;64(7):945–50. nested whole-brain analysis. Brain. 2005;128(Pt 1):213–26. 77. Andreae LC, Burrone J. Spontaneous neurotransmitter release shapes 54. Alexander AL, Lee JE, Lazar M, Field AS. Diffusion tensor imaging of the – dendritic arbors via long-range activation of NMDA receptors. Cell Rep. brain. Neurotherapeutics. 2007;4(3):316 29. 2015;10(6):873–82. 55. Alexander AL, Lee JE, Lazar M, Boudos R, DuBray MB, Oakes TR, et al. 78. Ye L, Allen WE, Thompson KR, Tian Q, Hsueh B, Ramakrishnan C, et al. Diffusion tensor imaging of the corpus callosum in autism. NeuroImage. Wiring and molecular features of prefrontal ensembles representing distinct – 2007;34(1):61 73. experiences. Cell. 2016;165(7):1776–88. 56. Travers BG, do PM T, Adluru N, Lange N, Destiche D, Ennis C, et al. Atypical 79. Jones DK. The effect of gradient sampling schemes on measures derived development of white matter microstructure of the corpus callosum in from diffusion tensor MRI: a Monte Carlo study. Magn Reson Med. 2004; males with autism: a longitudinal investigation. Mol Autism. 2015;6:15. 51(4):807–15. 57. Ellegood J, Babineau BA, Henkelman RM, Lerch JP, Crawley JN. Neuroanatomical analysis of the BTBR mouse model of autism using magnetic resonance imaging and diffusion tensor imaging. NeuroImage. 2013;70:288–300.